DPS Kubernetes monitoring

[Pedro_Marce]

Hi, 

 I am trying to get my head around of what I need to use as a way to calculate in DPS my consumption.

 All deployment is a set of kuberneted clusters deployed in GCP in GKE, currently monitored as full stack at host level (tried cloud native but we had lots of issues with logs and others).

 In DPS, our plan is to enable kubernetes monitoring with the new container monitoring, that should give us the very nice new kubernetes app and all monitoring of the cluster, but my understanding is that would not give us any application observability, for that we would need to deploy the operator in full stack or cloud native mode.

Ideally we would like to deploy it as cloud native, only enabling those namespaces where our code reside to limit costs, is this a good practice? Are we giving away any functionality by doing so? When we tried in the past logs were not ingested automatically and we had to ingest them via fluentd or external methods, while configuring it as full stack was processing correctly the logs.

How would we assess the costs, container monitoring would be clear, if we go full stack is clear too, all cluster nodes based on memory, with the cloud native option, every pod would be consider as a host as full stack? 

If anyone has a similar setup and could share their deployment would be very much appreciated to understand what would be the cost simulation moving to DPS.

Thanks,

[PacoPorro]

Hi 

Definitely, Cloud Native Full Stack is the best approach in your situation, logs can also be added using this deployment

Current limitations: https://docs.dynatrace.com/docs/shortlink/how-it-works-k8s#cloud-native-limitations

To see how to filter for selected namespaces, check:

https://docs.dynatrace.com/docs/shortlink/annotate

Regarding licensing please check https://docs.dynatrace.com/docs/shortlink/dps-containers, but I would recommend you to contact your Dynatrace sales team to help with this.

[Pedro_Marce]

Thanks for your reply.

Sales team redirected me here to clarify some questions, so it would help if someone would clarify.

Am I correct assuming that in that scenario I would be paying container monitoring based on the total number of pods in my cluster, plus full stack monitoring for the memory used in the namespaces I am monitoring?

Thanks,

[erh_inetum]

Hi Pedro,
Not exactly as far as I know.
You would pay container monitoring based on the pods you monitor, not for all.
When label "dynatrace.com/inject=true" is configured for specific namespaces, it enables deep monitoring for those namespaces and only the pods within these labeled namespaces will have the OneAgent injected and will be considered for billing based on the pod/hour usage.

Hope it helps.
Regards,
Elena.